|Home | About | Journals | Submit | Contact Us | Français|
Correspondence to: Frédérique Caillot, Inserm Unité 905, Faculté de Médecine-Pharmacie, 22 Bvd Gambetta, 76183 Rouen cedex, France. email@example.com
AIM: To search for transcription dysregulation that could (1) differentiate hepatocellular carcinoma (HCC)-free from HCC-related cirrhosis (2) differentiate HCC-free cirrhosis related to HCV from that related to alcohol intake.
METHODS: Using microarray analysis, we compared transcript levels in HCC-free cirrhosis (alcoholism: 7; hepatitis C: 7), HCC-associated cirrhosis (alcoholism: 10; hepatitis C: 10) and eight control livers. The identified transcripts were validated by qRT-PCR in an independent cohort of 45 samples (20 HCC-free cirrhosis; 15 HCC-associated cirrhosis and 10 control livers). We also confirmed our results by immunohistochemistry.
RESULTS: In HCC-free livers, we identified 70 transcripts which differentiated between alcoholic-related cirrhosis, HCV-related cirrhosis and control livers. They mainly corresponded to down-regulation. Dysregulation of Signal Transduction and Activator of Transcription-3 (STAT-3) was found along with related changes in STAT-3 targets which occurred in an etiology-dependent fashion in HCC-free cirrhosis. In contrast, in HCC, such transcription dysregulations were not observed.
CONCLUSION: We report that transcriptional dysregulations exist in HCC-free cirrhosis, are transiently observed prior to detectable HCC onset and may be appear like markers from cirrhosis to HCC transition.
Hepatocellular carcinoma (HCC) is the most prominent, primary liver cancer. Its main etiologies are viral hepatitis B or C (HBV; HCV), alcoholism or aflatoxin B1 intoxication, with both HCV and alcoholism currently increasing in incidence in Western countries and predominating as etiologies[1,2]. In most instances, HCC develops in the setting of chronic hepatitis and/or cirrhosis. Cirrhosis is the end stage of a chronic liver disease which results in regenerating nodules surrounding by fibrous septa and, ultimately, may lead to cancerous nodules. HCC has a poor prognosis but, apart from surgery, no major improvements in disease therapies have been recently reported, most likely because the heterogeneity of the disease and its various etiologies prevent any progress in our understanding of HCC development and mechanisms.
Numerous genome-wide analyses of abnormal gene expression in HCC as compared to normal, control livers, have resulted in identification of gene sets with altered expression[2,6,7], part of which result from underlying gene mutations and/or chromosome alterations[8–10] and account for a limited number of altered pathways[9,11]. A few similar studies have been done in HCC-free cirrhosis and they mostly considered markers for a pre-HCC condition, or selected pathways, or mixed etiologies[12,14]. In fact, the number of comparative studies devoted to HCC etiology has remained scarce, whether this was done in a clinical setting[15–19], cell lines or animal models with oncogene overexpression. In particular, the viral etiologies have been considered[17,18,20,22] whereas abnormal gene expression in alcoholism-dependent HCC has received very little attention. Therefore, the impact of etiology still remains an important issue[7,23]. We recently reported that a number of genome-wide abnormalities in alcoholism-associated vs HCV-associated HCC are etiology-dependent and some of them are of pathological relevance. Remarkably, the abnormal transcription levels that differentiate HCC nodules in an alcoholism-dependent vs HCV-dependent fashion can no longer discriminate between both etiologies when transcripts are measured in the surrounding cirrhosis. Yet, any etiology-dependent abnormalities that could be observed in HCC-free cirrhosis would be of interest. We investigated whether some transcription dysregulations could be found in HCC-free cirrhosis in an etiology-dependent fashion. Furthermore, we searched and found transcript dysregulations that differentiate HCC-free cirrhosis from peritumoral cirrhosis. We now report that such transcription dysregulations do exist in HCC-free cirrhosis and are observed prior to detectable HCC onset.
Non-alcoholic steatohepatitis, primary biliary cirrhosis and infant biliary atresia were excluded from this study. Chronic alcohol abuse was estimated as detailed. HBV and HCV infections were serologically determined in every patient and any HBV-positive patient was excluded. Patients with an HCC-free or HCC-associated cirrhosis were histologically diagnosed by trained pathologists (AF, PBS). Liver fragments came to our laboratory from the digestive surgery unit of Charles Nicolle Hospital (Rouen, France) or the pathology unit of Pellegrin Hospital (Bordeaux, France) under strict anonymity. HCC-free, cirrhotic tissue was obtained from transplanted patients. Peri-tumoral, cirrhotic tissue was taken at a distance from HCC resection whenever the latter was excised for curative purposes. Control, non-cirrhotic human liver (CL) was obtained from patients operated on for a benign liver tumor or metastasis of a non-hepatic cancer. According to the current French rules and ethical guidelines, neither informed consent nor advice from an ethical committee were requested prior to RNA analysis in tissues that would otherwise be disposed of. Various clinical features in a total of 69 cirrhosis without HCC, (n = 34) or with HCC (n = 35), as well as in a set of 18 histologically normal CLs are summarized in Table Table1.1. A METAVIR score from the combined extents of inflammation (A0-A3) and fibrosis (F0-F4) were histologically diagnosed by trained pathologists (AF, PBS).
RNA extraction from tissues stored at -80°C was done with Trizol. Our set of human cDNA probes dubbed Liverpool and tailored to a complete coverage of the human liver transcriptome under healthy or pathological conditions (ca. 104 genes), the associated LiverTools database, as well as the procedures from array preparation to data handling have all been detailed. In brief, every RNA sample was subjected to three rounds of hybridization and the resulting signals were normalized from the average signal of every spot (mean grey) on the matching hybridization image. The mean signal per transcript was used for selections of significantly regulated transcripts. Probe re-sequencing was done with an ABI3100 capillary sequencer (Applied Biosystems, Foster City, USA). Real-time qRT-PCRs of transcripts were done with a Light Cycler (Roche Diagnostics, Manheim, Germany). Transcript normalization was done with the 18S RNA. The primers designed with the Primer3 software (http://frodo.wi.mit.edu) are listed in a Table Table22.
Our raw data are deposited in the GEO repository under accession number GSE10356. The TIGR Multiexperiment viewer (Tmev version 2.2, http://www.tm4.org) was used for (1) unsupervised hierarchical clustering (UHC) using the Manhattan distance and complete linkage options; (2) supervised analyses such as the t-test or ANOVA adjusted with Bonferroni’s correction or K-nearest neighbour classification (KNNC) and (3) evaluation of sample re-assignment by a random procedure (jackknife-106 iterations). Another, supervised classification was done by Support Vector Machine (SVM) (http://svm.sdsc.edu/). The Gene Ontology Tree Machine (GOTM) program (http://bioinfo.vanderbilt.edu/gotm/) was used to categorize protein function(s) by ontology. Detailed protein functions were retrieved with the SOURCE (http://genome-www5.stanford.edu/cgi-bin/source/sourceSearch) and/or OMIM (http://www.ncbi.nlm.nih.gov/sites/entrez) tools. Protein networks were identified with Bibliosphere (www.genomatix.de). Statistics were carried out with the GraphPad Instat software, version 3 (http://www.graphpad.com/).
D4 zinc and double PHD fingers family 2 (DPF2) and plasminogen (PLG) protein levels were assessed in formaldehyde-fixed, paraffin-embedded, 5 mm thick liver section samples giving a total of 40 other cirrhosis without or with HCC (10 alcoholic and 10 HCV in each group), as well as a set of 10 other histologically normal CLs. This assessment was effected by immunohistochemistry using the ultraViewTM Universal DAB Detection Kit following the manufacturers instructions (Ventana Medical systems) with mouse anti-DPF2 IgGs (Abnova corporation) at 1 μg/mL or mouse anti-PLG IgGs (Interchim) at 38 μg/mL.
The percentage of positive cells was evaluated on the whole surface of the histological section and the staining intensity was estimated. Two scores from 0 to 4 were given as two independent visual scores by a trained pathologist and were evaluated using a LEICA DMR microscope equipped with a camera. The number of positive cells or P score was: 0, no positivity; 1, < 25%; 2, 25%-50%; 3, 50%-75% and 4, > 75%-100%. The determination of immunostaining intensity or I score was: 0, no staining; 1, very weak staining only seen at magnification × 10; 2, staining obviously seen at magnification × 10; 3, moderate staining seen at magnification × 2.5; 4, strong staining seen at magnification × 2.5. The IP score was obtained from the additional combination of the two parameters I + P.
First, with microarray data from 14 HCC-free cirrhosis samples (A1-A7, V8-V14) and eight CLs (CL1-CL8), we identified 30 transcripts whose levels differed between [A + V] cirrhosis vs CLs (t-test adjusted by Bonferroni’s correction, P < 0.05, Figure Figure1A).1A). In contrast, similar data from 20 peritumoral cirrhosis samples (PA1-PA10, PV11-PV20) and eight CLs identified only three transcripts, but they did not distinguish [PA-PV] from CLs (t-test adjusted by Bonferroni’s correction, P < 0.05, Figure Figure1B).1B). This HCC-dependent difference in transcript number was significant (30 vs 3, Fisher’s test, P < 0.0001). Furthermore, the expression levels of the 30 transcripts, which distinguished HCC-free cirrhosis, from CLs, did not distinguish [PA-PV] from CLs (data not shown). Thus, we identified transcripts which were dysregulated in HCC-free cirrhosis but not in peritumoral cirrhosis.
Next, the comparison of transcript levels in alcoholic HCC-free cirrhosis vs CLs identified 10 dysregulated transcripts (t-test adjusted by Bonferroni’s correction, P < 0.05). Likewise, we identified 49 dysregulated transcripts in HCV HCC-free cirrhosis vs CLs. We also found 33 transcripts that were differentially expressed when directly comparing alcoholic vs HCV HCC-free cirrhosis. Overall, we obtained a non-redundant list of 70 transcripts whose levels were able to completely distinguish between the three groups by UHC: alcoholic, HCV HCC-free cirrhosis and CLs (Figure (Figure2A).2A). This was supported by a jackknife procedure (100% success). In contrast, in HCC five transcripts failed to properly distinguish between these three groups (t-test adjusted by Bonferroni’s correction, P < 0.05, Figure Figure2B).2B). This HCC-related difference was significant (70 vs 5, Fisher’s test, P < 0.0001). Furthermore, the expression levels of these 70 transcripts, which separated alcoholic, HCV and CLs in an HCC-free context, did not distinguish between them in an HCC context (data not shown).
From our list of 70 transcripts obtained from our microarray data, we measured by real-time qRT-PCR the 20 most discriminant transcripts (as listed in Table Table3,3, footnote 2) which were differentially expressed according to etiology. Using both the 22 training samples (A1-A7, V8-V14 and CL1-CL8) and 30 further independent test samples (A15-A24, V25-V34 and CL9-CL18) in order to classify HCC-free cirrhosis by unsupervised (UHC) or supervised training/testing procedures (KNNC and SVM),we found that these 20 transcripts resulted in a classification accuracy of 79%-83% test samples (Table (Table33).
The abnormalities in transcript levels seen in HCC-free cirrhosis, but not in peritumoral cirrhosis, were further evaluated timewise. As shown in Figure Figure3A3A (upper left star) the expression levels in CLs (n = 18), HCC-free cirrhosis (n = 34) and peritumoral cirrhosis (n = 35) were measured by qRT-PCR. When comparing the above 20 transcript levels between HCC-free vs CLs, their mean level in HCC-free cirrhosis was up-regulated (TIMP1), unchanged (3/20 transcripts, 15%, DPF2, IFI27, STAT3), or mostly down-regulated (16/20, 80%). In contrast, this down-regulation was not found when cirrhosis was associated with HCC. Indeed, in peri-tumoral cirrhosis a significant return to the CL level or even an up-regulation was observed (Figure (Figure3A,3A, upper right star). Moreover, as shown in Figure Figure3B,3B, 9/20 (45%) transcript levels further displayed etiology-dependent differences found (1) only in HCC-free cirrhosis (alcoholism vs HCV, lower left star, 6/9 transcripts, 66%), or (2) only in peritumoral cirrhosis (lower right star, 2/9 transcripts, 22%, ARID1A, ORM2), or (3) in both (IFI27). Overall, these transcript dysregulations were mostly seen in HCC-free cirrhosis, often resulted from a transient down-regulation, and half of them were etiology-dependent in agreement with the initial selection.
Among the nine genes mentioned above which displayed etiology-dependent differences, DPF2 and PLG, whose antibodies were marketed for immunohistochemistry use, were selected. We quantified their protein levels in a total of 40 other cirrhosis without or with HCC (10 alcoholic and 10 HCV in each group), as well as in a set of 10 other histologically normal CLs.
The DPF2 protein level was significantly higher in HCC-free cirrhosis, and then decreased in HCC-associated cirrhosis to return to a level similar to that observed in CLs, but this level regulation was mild (data not shown).
The PLG protein level was also significantly different in CLs, HCC-free and HCC-associated cirrhosis. Indeed, the PLG level was significantly lower in HCC-free cirrhosis as compared to that observed in CLs and then increased in HCC-associated cirrhosis to return to a level quite similar to that observed in CLs (Figure (Figure4A).4A). The immunohistochemical pattern for CLs with a strong hepatocellular staining was shown in Figure Figure4D).4D). The PLG protein level also displayed etiology-dependent differences. Indeed, the decrease of the staining was significantly higher in HCC-free HCV cirrhosis (Figure (Figure4E)4E) than in HCC-free alcoholic cirrhosis (Figure (Figure4B)4B) and in HCC-associated cirrhosis the increase with a return to the baseline was the same whatever the etiology (HCV or alcohol) (Figure (Figure4C4C and andF).F). Thus, the PLG protein level confirmed our results obtained at the transcriptional level.
We first investigated whether transcript variations could point to functional dysregulation in HCC-free cirrhosis. By comparing our list of 70 transcripts and our Liverpool, a different frequency of dysregulated transcripts in eight functional subsets was found (detailed as a Table Table4):4): (1) cell proliferation (P = 0.01); (2) regulation of cell migration (P = 0.009); (3) blood vessel development (P = 0.007); (4) lipid metabolism (P = 0.007); (5) antigen processing and presentation (P = 0.001); (6) acute inflammatory response (P = 0.002); (7) NADH dehydrogenase activity (P = 0.005) and (8) oxidoreductase activity (P = 0.001).
Within our set of 70 transcripts, 23 transcripts with available information from Bibliosphere exhibited an etiology-associated level variation in HCC-free cirrhosis (Figure (Figure5),5), but not in peritumoral cirrhosis (data not shown). This resulted from a variable extent of down-regulation (17/23 transcripts, 74%) in a single etiology or both. In turn, this down-regulation resulted, at least partly, from variable, etiology-dependent regulation of STAT-3 and its target genes (lower right area of Figure Figure55).
Our search first focused on the significance of dysregulations found in HCC-free cirrhosis. We investigated the influence of an HCC-free vs peritumoral environment of cirrhosis and we searched for early markers of HCC onset. The ideal study design would be to compare cirrhosis that develops into HCC and cirrhosis that does not develop into HCC in the follow-up of the disease, but such samples were very scarce because the follow-up among cirrhotic patients would be extremely long and difficult. So, we compared HCC-free and HCC-associated cirrhosis and we used a stringent selection of informative transcripts because sharp, timewise variations of potential markers of HCC occurrence were to be identified. We have shown that transcription dysregulation does exist in HCC-free cirrhosis. This is observed before any histologically detectable HCC nodule is seen, and hence supports the “field cancerization” model. Specifically, from our GOTM results, it seems plausible that malignant transformation of cirrhosis could be favored by abnormal expression of factors regulating cell migration, cell proliferation and blood vessel development. These dysregulations in cirrhosis mainly correspond to down-regulations, and they usually re-normalize or are even up-regulated in peritumoral cirrhosis. It remains to be determined how such transient down-regulations, which have been as yet unreported, contribute to HCC initiation.
We next carried out a similar search, further integrating alcoholism and HCV etiologies. A few, etiology-dependent markers for a high vs low risk of HCC development have been previously identified by others, but they relied on an unproven assignment of alcoholism or HCV patients to either risk group and they could not predict HCC occurrence in a timewise fashion. We had previously reported that abnormalities of some transcript levels are observed to a different extent in HCC developed on alcoholic-associated cirrhosis vs HCV-associated cirrhosis, whereas they remain similar in peritumoral cirrhosis, thus indicating that these abnormalities are etiology-related in HCC tumors only. In the same way, in the present study, we found transcript dysregulation only in HCC-free cirrhosis and not in peritumoral cirrhosis. We now document that histochemical evaluation of the PLG protein level confirms our results obtained at the transcriptional level. In contrast, for DPF2, dysregulation observed at both transcript and protein levels were in opposite directions, but discrepancies between transcript levels and protein levels have been previously noticed[26,27]. As markers of HCC occurrence are still very scarce[4,28], our observation on transcripts and proteins is of strong interest in early HCC diagnosis. This will need to be further evaluated in HCC-free/cirrhosis-free, fibrotic livers, with selection of marker combinations.
Some up- or down-regulations of transcription factors in HCC have already been documented, but the facts that dysregulation of STAT-3 and a related gene network take place in HCC-free cirrhosis, and in an etiology-dependent fashion, are novel findings. The JAK-STAT pathway is critical in the proinflammatory cytokine-driven inflammatory response provided by hepatocytes and it is tempting to speculate that a weakening of this defense in early cirrhosis may participate in HCC development. However, this mechanism appears unlikely. Indeed, our data were obtained by comparison of alcoholic vs HCV cirrhosis samples whose extent of inflammation was similar and still had different STAT-3 regulation. The HCV has a clear effect on the activity of STAT-3, but the meaning of this is controversial. Some studies show inhibition of STAT-3 activity while others show activation of STAT-3[32–34]. Our data are in keeping with documented HCV proteins/STAT-3 interferences and STAT-3 activation in HCV-induced liver disease. STAT-3 directly affects cell proliferation, cell differentiation and angiogenesis. Moreover, STAT-3 and its targets are regulated in some cancers, such as breast and prostate cancer. Thus, the dysregulation of the STAT-3 pathway which follows HCV infection may participate in HCC development at an early stage of hepatocyte dysplasia. In addition, recent reports have highlighted the potential of STAT-3 as a therapeutic target in different neoplasms[36,37].
In conclusion, our data point to major transcription dysregulations in HCC-free cirrhosis. These dysregulations often result from a transient dysregulation, and half are etiology-dependent. Our observations open new avenues for the follow-up of HCC-free cirrhosis because dysregulated transcripts or proteins may be appear like markers for the cirrhosis to HCC transition. In order to complement these results, studies performed at an earlier state before cirrhosis, i.e. on fibrosis samples are now under investigation.
Chronic viral hepatitis C (HCV) infection and alcoholism are two important causes of cirrhosis and hepatocellular carcinoma (HCC). Liver transcriptome analysis has resulted in the identification of genes with an aberrant expression according to different pathophysiological states. In the present work, we investigated whether some transcription dysregulations could be found in HCC-free cirrhosis in an etiology-dependent fashion. Furthermore, we searched and found transcription dysregulations that differentiate HCC-free cirrhosis from peritumoral cirrhosis.
Numerous, genome-wide analyses of abnormal gene expression in HCC as compared to normal, control liver, have resulted in identification of gene sets with altered expression. Few similar studies have been done in HCC-free cirrhosis. In fact, viral etiologies have often been considered whereas abnormal gene expression in alcoholism-dependent HCC has received very little attention. Therefore, the impact of etiology still remains an important issue. We recently reported that a number of genome-wide abnormalities in alcoholism-associated vs HCV-associated HCC are etiology-dependent and some of them are of pathological relevance. Remarkably, the abnormal transcript levels that differentiate HCC nodules in an alcoholism-dependent vs HCV-dependent HCC can no longer discriminate between the two etiologies when transcripts are measured in the surrounding cirrhosis. Yet, any etiology-dependent abnormalities that could be observed in HCC-free cirrhosis would be of interest.
These data point to major transient transcription dysregulations in HCC-free cirrhosis. These observations open new avenues for the follow-up of HCC-free cirrhosis because dysregulated transcripts or proteins may be appear like markers for the cirrhosis to HCC transition.
The aims of the study were to identify genes that were differentially expressed between HCC-free and HCC-related cirrhosis. The differentially expressed genes were further investigated to see if they were associated with alcoholism and HCV etiologies. The authors suggested that genes that were deregulated in HCC-free cirrhosis might serve as markers for the cirrhosis to HCC transition.
We thank Parey Simon for his excellent technical assistance for immunohistochemistry procedures. F.C. is recipient of a fellowship from the French Ministry for Research and A.R.C. C.D. is the recipient of a fellowship from Ligue contre le Cancer.
Supported by In part by Grants from A.R.C., Ligue contre le Cancer, I.R.E.B. and Conseil Régional de Haute-Normandie to J.P.S.
Peer reviewer: John M Luk, Associate Professor, Department of Surgery, The University of Hong Kong, 21 Sassoon Road, Pokfulam, Hong Kong, China
S- Editor Tian L L- Editor Kerr C E- Editor Zheng XM